Re: repodata - xz
On Sun, Jun 22, 2014 at 9:57 PM, Jon <jdisnard(a)gmail.com> wrote:
On Sun, Jun 22, 2014 at 6:26 AM, drago01 <drago01(a)gmail.com>
wrote:
> On Sun, Jun 22, 2014 at 8:11 AM, Tim Lauridsen <tim.lauridsen(a)gmail.com>
wrote:
>>
>> On Sat, Jun 21, 2014 at 8:28 PM, Jon <jdisnard(a)gmail.com> wrote:
>>>
>>> Would it be too much trouble to use the sqlite data in DNF?
>>> I suppose it would be a step backwards to have our primary (future)
>>> tool using gzip metadata.
>>
>>
>> the information in the sqlite files is orded i a way the yum is using the
>> data, dnf parses the .xml files into another format used by libsolv.
>> the sqlite was invented because the yum .xml to sqlite format was slow and
>> has to happen on every client out there, so it was moved serverside to speed
>> up things.
>
> Well can't we change create repo to create xml.xz files (in addition
> to .gz to preserve backwards compatibility) ?
> --
Perhaps we should make createrepo default to XZ in Fedora rawhide, and
make all the tools do the right thing?
Would be nice if createrepo was influenced by a system-wide config
file, so we could easily implement that idea.
Otherwise we are forced to work around createrepo in our release engineering.
Regardless we should strive to make all repo metadata XZ compressed,
unless there is a good reason not to do that?
Well even if we do that we need to support the old format (i.e fall
back when xz is not available) otherwise old (third party / custom)
repos would stop working.